Investigate with memory profiling
See if it is possible to adapt work done here for ruby-prof / kcachegrind to pyprof2calltree:
The goal would be to track function / method calls that eat a lot of memory in the call tree.
This issue is yet unassigned because I don't know whether I will have the time to do it myself. So please do not hesitate to clone pyprof2calltree and hack around to add memory profiling if you feel so.